Research On Accurate Auxiliary Diagnosis Of Lung Cancer Based On Intracellular Fluorescent Fingerprint Information

At present,the morbidity and mortality of lung cancer are increasing all over the world,and lung cancer has become one of the malignant tumors that threaten the health and life.The differences in benign and malignant lung cancer,pathological types and genotypes directly affect the choice of clinical treatment and clinical prognosis.The precise diagnosis of different types of lung cancer has a profound impact on clinical treatment.At present,the clinical diagnosis of lung cancer mainly relies on imaging examinations,pathological examinations and other methods.However,these methods are greatly influenced by the subjective consciousness and experience of doctors.In the diagnosis of molecular typing of lung cancer,the mainstream method of diagnosis is DNA sequencing,but there are also disadvantages such as high inspection cost and long time.Therefore,the method of using image processing technology to assist in the diagnosis of lung cancer came into being,but when the morphological features are ambiguous,the deep learning model has a low effect on the auxiliary diagnosis accuracy of general medical samples.Compared with traditional image samples,the intracellular fluorescence fingerprint information of hyperspectral imaging includes spectral information in addition to spatial structure information,which provides a new sample type and approach for lung cancer classification tasks.On this basis,this paper analyzes and studies the classification of benign and malignant lung cancer,lung cancer EGFR gene mutation and lung cancer pathological types based on intracellular fluorescent fingerprint information.The research thesis can be roughly divided into three parts.First,the classification of benign and malignant lung cancer is studied.The random forest algorithm and one-dimensional convolutional neural network are trained with one-dimensional data.Residual networks with three depths are trained with two-dimensional data.The algorithm aims to determine the model with the best classification performance by comparing the classification results of multiple models.Subsequently,the above model was used to classify whether EGFR gene mutation occurred in lung cancer,and its one-dimensional model and two-dimensional model were compared and analyzed.Finally,this paper discusses the classification of lung cancer pathological types,compares various deep learning models,and proposes an optimization method based on ensemble learning according to the classification of the models,and finally forms an ensemble learning model with the best classification performance.The experimental results show that because the two-dimensional image data contains more structural information of fluorescent images,the classification performance of the residual network model is better than that of the one-dimensional model in the study of benign and malignant lung cancer classification.The optimal model is verified by a 50-layer residual network.The AUC area of the set is 0.998.In addition,the intracellular fluorescent fingerprint information can better display the difference in the characteristics of lung cancer with or without EGFR gene mutation,which improves the accuracy of the classification model to a certain extent.Make the AUC area of the validation set for the optimal model to be 0.998.In the research on the classification of lung cancer pathological types,the classification effect of the two-dimensional model is slightly worse than that of the one-dimensional model,but the three residual network models with different depths have their own advantages in classification performance on different types of samples.After the model is optimized,the macro-average AUC area of the final ensemble learning model is 0.97,and the micro-average AUC area is 0.96.The ensemble learning method effectively improves the classification performance of the model.Through the research of this paper,it is found that the effect of intracellular fluorescent fingerprint information in the accurate diagnosis of lung cancer can better meet the clinical needs,and can help doctors in clinical decision-making.The application of intracellular fluorescent fingerprint information can provide accurate clinical diagnosis of tumors.New ideas and new solutions.